This invention relates generally to communication networks and, more particularly, relates to a method of adapting video packet generation and transmission rates to available resources in a communications network.
In communication networks, the communication channel is a precious resource that needs to be shared intelligently between multiple communication sources. To efficiently utilize this resource, an appropriate channel access scheme must be selected. The target application and the corresponding underlying traffic that is envisioned to traverse the communication network largely influence this selection. Typically, the underlying traffic is envisioned to be integrated packet voice and data communications. Accordingly, currently utilized channel access schemes are biased towards supporting integrated packet voice and data communications while packet video communication is generally ignored.
For supporting integrated packet voice and data communications, several multiple access schemes have been proposed in the prior art. Specifically, these schemes can be organized into three well known categories, namely, Frequency Division Multiple Access (FDMA), Time Division Multiple Access (TDMA), and Code Division Multiple Access (CDMA). Among these three schemes, schemes based on TDMA, where time is divided into frames and frames are divided into slots, have enjoyed the most acceptances.
Generally, in TDMA schemes, a communication source transmits a transmission packet over the communication channel upon the commencement of its assigned time slot. A network router, in the form of a base station, server, or the like, receives the transmission packet and then assists in routing the transmission packet towards its final destination. Depending on which time-slots are assigned to the communication source, TDMA schemes are classified into two types: basic TDMA and dynamic TDMA. In basic TDMA, specific time-slots are assigned to the communication source for the entire duration of the connection. In contrast, in dynamic TDMA, the specific time-slots assigned to the communication source can vary during the lifetime of the connection. Dynamic TDMA schemes are essentially a compromise between random access and controlled access type protocols. These schemes contain at least one contention phase in which new communication sources attempt to announce their presence by transmitting connection establishment request messages to the network router. Examples of such TDMA based schemes include Packet Reservation Multiple Access (PRMA), Reservation-ALOHA and Reservation-MA (R-MA), Dynamic Reservation Multiple Access (DRMA), and Dynamic-Time Division Multiple Access (D-TDMA). The frame structures used in PRMA, R-MA, DRMA, and D-TDMA are each illustrated in prior art FIG. 1.
Turning first to the frame structure utilized with PRMA, a slot S within a frame F is either available A or reserved R. Both voice and data communication sources contend for the available slots according to the voice and data transmission probabilities that are set during the system design. If a voice communication source succeeds during the contention phase, an available slot is assigned to that communication source and is labeled as reserved. The reserved slot is thus made available to that communication source in subsequent frames during the time it is actively generating and transmitting voice packets. When the voice communication source has no more voice packets to transmit, it looses its reservation and goes back to the contention phase when it has additional voice packets to transmit. For the case of pure data communications, if a data communication source succeeds during the contention phase, it uses the available slot to transmit the data packet. However, this slot is not reserved in subsequent frames and remains available to be contended for in the immediately following frames.
Like PRMA, R-MA allows multiplexing to be performed at the talkspurt level and a voice communication source keeps a slot for as long as it is active while a data communication source must contend for a slot during each frame. However, in contrast to PRMA, R-MA requires that some amount of bandwidth be kept available for use in servicing connection requests. This bandwidth is provided in the form of dedicated contention slots C that are further divided into a plurality of mini-slots MS. Thus, in R-MA, it is on the mini-slot boundaries that connection requests are made in accordance with permission probabilities.
In DRMA, similar to PRMA and unlike R-MA, each available slot can be used for information transmission or for channel reservation and no slots are dedicated for servicing connection requests. Furthermore, similar to R-MA and unlike PRMA, when serving as a contention slot, a slot is divided into a plurality of mini-slots on whose boundary connection requests are made in accordance with permission probabilities. Again, once a slot is reserved for a voice communication source, it can be used by that voice communication source in subsequent frames for as long as there are voice packets to transmit. Data communication sources are assigned slots in only one frame for data packet transmissions.
Finally, in D-TDMA frames are further divided into contention slots, voice slots, and data slots. Voice communication sources are allocated slots from the voice slots and data communication sources are allocated slots from the data slots. A registered voice communication source is assigned a voice slot that is maintained until no further voice packets are transmitted. Data communication sources are again assigned slots in only one frame. Furthermore, like R-MA and DRMA, connection requests are made on mini-slot boundaries.
While these discussed schemes do provide respectable quality for voice and data communications, they nevertheless tend to fall short when evaluated for real-time video communications. For example, the described protocols fail to provide any mechanism for guaranteeing sustained bandwidth, bounded delay, and, accordingly, quality of service guarantees for video communications. Quality of service is an essential ingredient for the success of many real-time video applications and without it, under heavy loads, video tends to exhibit poor and sometimes intolerable quality.
A further disadvantage resides in the fact that the frame length in the described schemes is typically designed to be equal to the packet generation period of the voice encoder. In this manner, since one voice packet will be generated in one frame time, both delay and buffering are bounded for voice transmissions. For video transmissions, this choice of frame length has no meaning since video encoders typically generate packets at rates generally faster than voice encoders do. Accordingly, the one slot per frame guarantee does not function to prevent excessive delays or overflow buffering for video transmissions again resulting in video quality degradation.
Yet another disadvantage resides in the fact that these schemes typically omit video communication sources as a separately identifiable communication source when assigning slots on a priority basis. Accordingly, when video packets are treated as voice packets, the typically higher priority given to voice packets coupled with the high demand for bandwidth required by video transmissions tends to overwhelm the resources of the communication network while degrading all on-going connections. Similarly, when video packets are treated as data packets, the data packet requirement of contending for every slot tends to result in frequent collisions causing excessive delays in video transmissions that again function to lower both the quality of on-going connections and the overall bandwidth utilization of the communication network.
Finally, the described schemes that rely completely on contention to determine slot allocation will perform poorly under heavy load. Accordingly, when video communications are introduced into the communication network, the amount of data in the system is increased to the point where collisions are bound to escalate. This results in excessive delays and packet dropping for on-going video connections. This occurs even in those schemes that reserve a fixed amount of resources for contention purposes such as R-MA and D-TDMA.
From the foregoing, it is seen that a need exists for an improved channel access protocol. In particular, such a protocol is needed for use in establishing a full service network that provides comprehensive support for integrated transport of voice, video and data communications.
According to this need, the present invention is generally directed to a method for allocating between multiple communication sources a communication channel in a communication network. The method is performed by dividing the communication channel into a plurality of frames, dividing each of the frames into a plurality of slots, and dividing some of the plurality of slots into a plurality of first mini-slots. The mini-slots are provided for use by the multiple communication sources to request the establishment of a new voice, data, or video transmission connection over the communication channel. Additionally, one of the plurality of slots is divided into a plurality of second mini-slots for use by the multiple communication sources to request the establishment of a new voice, data, or video transmission connection over the communication channel and for use by the multiple communication sources to augment an existing video connection over the communication channel. In this manner a dynamic Time Division Multiple Access protocol frame is created that is available to support the integrated transport of voice, video and data communications over the communications network.
The subject invention is also described in P. Bahl, xe2x80x9cARMAPxe2x80x94An Energy Conserving Protocol for Wireless Multimedia Communications,xe2x80x9d IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, Boston, Mass., (Sep. 8-11, 1998) which is incorporated herein by reference in its entirety.